Csaba Ringhofer created IMPALA-12902:
----------------------------------------
Summary: Event replication is can be broken if
hms_event_incremental_refresh_transactional_table=false
Key: IMPALA-12902
URL: https://issues.apache.org/jira/browse/IMPALA-12902
Project: IMPALA
Issue Type: Bug
Components: Catalog
Reporter: Csaba Ringhofer
when setting hms_event_incremental_refresh_transactional_table=false
metadata.test_event_processing.TestEventProcessing.test_event_based_replication
fails at the following assert:
[https://github.com/apache/impala/blob/6c0c26146d956ad771cee27283c1371b9c23adce/tests/metadata/test_event_processing_base.py#L234]
Based on the logs catalogd only sees alter_database and transaction events in
this case, so if the transaction events (COMMIT_TXN) are ignore, then it
doesn't detect the change in the table.
This seems strange as the commit that added the test is older than the one that
added hms_event_incremental_refresh_transactional_table
[https://github.com/apache/impala/commit/e53d649f8a88f42a70237fe7c2663baa126fed1a]
vs
[https://github.com/apache/impala/commit/097b10104f23e0927d5b21b43a79f6cc10425f59]
So it is not clear to me how could the test pass originally. One possibility is
that different events were generated in HMS at that time.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
